Hi!

I’m interested in using OpenMC to determine the dose rate outside a spherical Am241 source with a radius of 1 cm. I set up the spherical spatial distribution of the source as follows:

```
sphere_radius = 1
r = openmc.stats.PowerLaw(0, sphere_radius, 1)
theta = openmc.stats.Uniform(0, pi)
phi = openmc.stats.Uniform(0, 2*pi)
spatial_dist = openmc.stats.SphericalIndependent(r, theta, phi)
source.space = spatial_dist
```

Originally, I used a source energy of 60 keV (the main line from Am241), but got results that seemed a bit to low when comparing with references. To cross-check the method, I increased the source energy to 1 MeV (still sampling from the spherical distribution), and implemented the same geometry and source in Serpent and MCNP for benchmarking. I tally the photon current going out of the 1-cm sphere. The results are shown below (the current across the sphere surface per initial source particle):

```
Spherical source:
OpenMC: 0.5898 (current in 1 MeV bin: 0.4694)
Serpent: 0.6218 (current in 1 MeV bin: 0.5011)
MCNP: 0.6213 (current in 1 MeV bin: 0.5012)
```

Clearly, OpenMC yields both a lower total current and a lower fraction of the uncollided 1-MeV photons passing through the sphere surface than Serpent and MCNP. Serpent and MCNP are in good agreement.

To investigate further, I then sample source particles only at the sphere centre (i.e. a point source). I tally the current passing through the sphere surface, and get the following results:

```
Point source:
OpenMC: 0.4613 (current in 1 MeV bin: 0.3344)
Serpent: 0.4648 (current in 1 MeV bin: 0.3345)
MCNP: 0.4653 (current in 1 MeV bin: 0.3348)
```

A much better agreement between all three codes is clearly seen. I attach a plot comparing the three spectra of the photons leaving the sphere (for both the point and the spherical source):

From the figure, the discrepancy in the 1-MeV peak (for the spherical source) is not really seen, but there are clearly some discrepancies between OpenMC and the other codes at lower energies (for both the point and the spherical source), perhaps due to differences in how the photon physics are handled? As can be seen, Serpent and MCNP are generally in good agreement.

What I cannot understand is why there is such a big difference in the tally results when using the spherical source distribution - a difference which is not observed when using the point source. This makes me think that there could be something going on with the sampling of the spherical source distribution in OpenMC.

Do you have any idea what could be the issue here? I’m using OpenMC version 0.13.0.

Thanks in advance!